OpenAI spent $34B in 2025, still burned cash before IPO

Key Takeaways

• OpenAI spent $34 billion in 2025, with $19 billion on R&D and $6 billion on sales and marketing
• Despite $13 billion in revenue exceeding targets, the company still operates at a significant loss
• The company is targeting a $1 trillion valuation for its planned IPO

OpenAI burned through $34 billion last year. The ChatGPT maker's audited financials, reported by the Financial Times, show the company spent more than double what it earned as it races to stay ahead in the AI arms race.

The breakdown: $19 billion went to research and development. Nearly $6 billion covered sales, marketing, and other operational costs. OpenAI pulled in $13 billion in revenue for 2025, beating its $10 billion target. But the math is stark. The company lost roughly $21 billion in a single year.

These figures arrive as OpenAI prepares for what could be the largest tech IPO in history. The company is targeting a $1 trillion valuation, a number that would place it among the most valuable companies on Earth before it has ever turned a profit.

Where did the $34 billion go?

Training frontier AI models is extraordinarily expensive. OpenAI's R&D spend of $19 billion reflects the cost of compute. GPUs from Nvidia do not come cheap, and training runs for models like GPT-5 require tens of thousands of them running for months.

The $6 billion on sales and marketing shows OpenAI's push to convert its viral ChatGPT success into enterprise revenue. The company has aggressively pursued corporate contracts, API partnerships, and consumer subscriptions to ChatGPT Plus.

“The intensity of the AI compute arms race is unlike anything we have seen in tech history; these numbers are not just costs, they are the ante to play.”

— Sarah Mukherjee, Lead Tech Analyst at Global Capital Research

That remaining $9 billion? Infrastructure, headcount, cloud credits, and the operational overhead of running one of the fastest-growing companies in tech history. OpenAI has hired aggressively from Google DeepMind, Meta AI, and other competitors.

Can a company losing $21 billion justify a $1 trillion valuation?

This is the question investors will have to answer. OpenAI is betting that its lead in generative AI translates into dominance across enterprise software, consumer applications, and eventually artificial general intelligence.

The bull case: OpenAI sits at the center of every major AI application being built today. Its API powers thousands of startups. ChatGPT has over 100 million weekly active users. If the company achieves meaningful progress toward AGI, today's losses become rounding errors.

The bear case: $21 billion in annual losses is unsustainable. Microsoft, OpenAI's largest backer, has already committed over $13 billion but has also hedged by building competing models. Competitors like Anthropic, Google, and Meta are closing the capability gap. And the compute costs show no sign of dropping.

The structural shift behind these numbers

OpenAI began as a nonprofit research lab. That mission has been steadily abandoned. The company restructured in 2019 to create a capped-profit entity, then again in recent years to remove those caps entirely. The IPO will complete the transformation into a conventional for-profit corporation.

This matters because it changes how we should evaluate the company. A nonprofit pursuing beneficial AI can justify losses indefinitely if donors keep funding the mission. A public company targeting $1 trillion needs a path to profitability that investors can model.

OpenAI's path depends on two bets: that compute costs will fall faster than revenue needs grow, and that its technology lead is durable. Neither is guaranteed.

How does this compare to other AI companies?

OpenAI's spending dwarfs its competitors. Anthropic raised $7 billion total. Google DeepMind operates within Alphabet's broader budget but has never disclosed standalone figures approaching this scale. Meta spent roughly $15 billion on Reality Labs in 2023, and that unit was widely criticized for excessive losses.

The closest comparison might be Amazon Web Services in its early years, or Tesla before profitability. Both companies burned cash for years before their bets paid off. Both also had skeptics who turned out to be wrong.

The difference: AWS and Tesla were building physical infrastructure and manufacturing capacity that created durable moats. Software can be replicated. Whether OpenAI's data, talent, and brand create a similar moat remains the central question.

What happens if the IPO falls short?

A $1 trillion valuation is not assured. Public markets have been brutal to unprofitable tech companies since 2022. If OpenAI prices below expectations, it could trigger a repricing across the entire AI sector.

The company has runway. Its last funding round reportedly valued it at $157 billion, and Microsoft's ongoing support provides both cash and compute. But at a $21 billion annual burn rate, even deep pockets have limits.

Frequently Asked Questions

How much money did OpenAI spend in 2025?

OpenAI spent $34 billion in 2025, according to audited financial figures reported by the Financial Times. This included $19 billion on R&D and $6 billion on sales and marketing.

What is OpenAI's target valuation for its IPO?

OpenAI is targeting a $1 trillion valuation for its planned IPO, which would make it one of the largest tech IPOs in history.

Is OpenAI profitable?

No. OpenAI generated $13 billion in revenue in 2025 but spent $34 billion, resulting in a net loss of approximately $21 billion.

Why does OpenAI spend so much on R&D?

Training frontier AI models requires massive compute resources, primarily expensive Nvidia GPUs running for months. The $19 billion R&D spend reflects these compute costs and the company's push to maintain its technology lead.

When will OpenAI go public?

The specific timing has not been announced, but the company is actively preparing for an IPO based on the release of audited financials.

Source: Tech-Economic Times / ET